FluxyChatEnterprise
MCP identity and audit
Register MCP servers, tool provenance, and review tool-call audit logs for agent security.
FluxyChat tracks MCP server identity, tool instructions/provenance, and an audit log for every tool call routed through the platform MCP layer.
Configure from Settings → MCP identity or the admin APIs below.
Register an MCP server
curl -X POST "$WORKER_URL/admin/mcp-identity/servers" \
-H "Authorization: Bearer $ADMIN_JWT" \
-H "Content-Type: application/json" \
-d '{
"name": "crm-tools",
"version": "2.1.0",
"vendor": "acme",
"instructions": "Use only for read-only CRM lookups."
}'Register tool provenance
Link a tool to a registered server with origin metadata (installed, builtin, remote):
curl -X POST "$WORKER_URL/admin/mcp-identity/tools" \
-H "Authorization: Bearer $ADMIN_JWT" \
-H "Content-Type: application/json" \
-d '{
"serverName": "crm-tools",
"toolName": "lookup_contact",
"instructions": "Requires customerId in args.",
"origin": "installed"
}'Registry and audit log
# Full registry + built-in serverInfo
curl "$WORKER_URL/admin/mcp-identity/registry" \
-H "Authorization: Bearer $ADMIN_JWT"
# Recent tool calls (default limit 50)
curl "$WORKER_URL/admin/mcp-identity/audit?limit=100" \
-H "Authorization: Bearer $ADMIN_JWT"Built-in FluxyChat MCP tools are logged automatically when invoked through the worker MCP handler.
Security notes
- Treat
instructionsas security policy — they are shown to agents and stored per tenant. - Audit entries retain server/tool names and outcome hashes, not full argument payloads (PII minimization).
- Pair with AI Governance for risk-tier evaluations on tools and models.